creating symbolic matrices is insanely slow

[williamstein]

On the fastest modern hardware we have:

sage: time m = matrix(SR, 20, [1..20^2])
CPU times: user 0.34 s, sys: 0.12 s, total: 0.45 s
Wall time: 1.05

which is frickin' slow. And the time isn't just in coercion, since

sage: time v = [SR(a) for a in [1..20^2]]
CPU times: user 0.01 s, sys: 0.00 s, total: 0.01 s
Wall time: 0.01

Component: calculus

Issue created by migration from https://trac.sagemath.org/ticket/2053

[williamstein]

comment:1

The attached patch fixes this problem. Now the following works and speeds
up the above benchmark by a huge amount:

sage: time m = matrix(SR, 20, [1..20^2])
CPU times: user 0.02 s, sys: 0.00 s, total: 0.02 s
Wall time: 0.07
sage: time m = matrix(SR, 50, [1..50^2])
CPU times: user 0.07 s, sys: 0.02 s, total: 0.09 s
Wall time: 0.46
sage: time m = matrix(SR, 100, [1..100^2])
CPU times: user 0.26 s, sys: 0.02 s, total: 0.27 s
Wall time: 0.49

Even a 500x500 matrix is possible, which used to be out
of the question.

sage: time m = matrix(SR, 500, [1..500^2])
CPU times: user 6.79 s, sys: 0.39 s, total: 7.17 s
Wall time: 13.32

[williamstein]

Attachment: trac-2053-symbolic_matrices_slow.patch.gz

[mwhansen]

comment:2

I cannot apply this due to a failed hunk with the docstring in set_from_list.

[williamstein]

Attachment: trac-2053-fix_symbolic.patch.gz

apply this INSTEAD!

[mwhansen]

comment:3

Looks good to me.

[sagetrac-mabshoff]

comment:4

Merged the second patch only in Sage 2.10.2.alpha2